One horse. One bet. Every day. Sly Man’s read on the day’s card — the shortlist, the price, and the case for the pick.
Browse the logbookWhen the scores are this tight, you stop trying to find daylight between them and you let the market do the separating — that is the read I keep coming back to today. The scan covered all 38 races declared on the 13th of October 2026, across 375 runners. Filters applied in sequence: an odds band of 2.0 – 50.0 decimal (no odds-on, nothing longer than 49/1), then a field-size gate of 6 – 14 runners to cut out the noise at either extreme, then a lay-of-day collision check that removed four runners flagged by the laying model. Feature weights ran as follows: market_move leads at 1.6, speed at 1.5, form at 1.2, class_change at 1.0, trainer_strike and going each at 0.8, and draw currently unavailable pending the v2 data layer.
The day opened with 375 runners across 38 races. The odds-band filter trimmed that to 317, the field-size gate brought it down to 190, and the lay-of-day collision check dropped a further four to leave 186 runners in contention for the selection. Those 186 form the pool the composite scores were built from.
Six horses made the shortlist. Koffee And Kale leads on composite at 6.96/10 — a genuine steamer at Wolverhampton 19:30, speed z-score of +1.82, and a +15.1% model edge over the market-implied probability. Carrhill sits second at 6.86/10, with the best speed reading on the board (z = +2.12) but three unavailable features and no going data whatsoever. Kenough is third at 6.82/10 and carries the longest price on the shortlist at 5/1 — a +108.4% model edge that looks extraordinary but rests on an almost entirely data-sparse profile (no form string, no class, no going history). Bradbury comes in at 6.74/10, another steamer at 5/1 with solid going data. Soft Landing scores 6.71/10 but carries a cold trainer (Saeed bin Suroor, 7% strike rate over 30 days) and a negative model edge of -1.7%. Zachary Hickes rounds out the six at 6.66/10. The top composite leads the sixth by just 0.30 points — well inside the 0.4-point threshold that would count as a clear lead. When the engine cannot separate them, the tie-break rule is the shortest price on the shortlist: the horse the market itself rates likeliest to win.
The read here is structural rather than handicapping: the scores are level, the engine's tie-break protocol points to the shortest price on the shortlist, and that horse is Zachary Hickes at 6/5 (decimal 2.2). I had a long look at whether I could argue past the rule and it did not survive contact with the data — Koffee And Kale is the top scorer and I am not pretending otherwise, but the margin is 0.30/10 and the protocol is explicit. The market is doing the separating today, and the market has made Zachary Hickes its firmest opinion on the shortlist. That carries weight when the model cannot.
The feature breakdown for Zachary Hickes is solid without being spectacular. Speed comes in at a z-score of +1.17 (SR 87 vs a field mean of 68.3) — not the sharpest on the board but enough to lead this particular Class 4 field at Wolverhampton over 1m5f219y. Form reads '/64123' with a decay-weighted average of +0.79 on the latest positive position — a consistent upward trajectory over the recent sequence. Going evidence is limited but present: 1-from-4 (2 placed) on good or adjacent ground, which is at least a data point where several others on the shortlist have none at all. Roger Varian's strike rate of 10/70 = 14% over the last 30 days is on par for the grade. And the market move is a steamer — a strict majority of bookies shortened, none drifted — earning the maximum +2.00 on the market_move feature, the highest-weighted input in the model at 1.6.
The one number that gives me honest pause is the market-implied edge: the model estimates a 34.0% win probability against the market's 40.6% implied, a negative edge of -25.3%. That means the market is pricing Zachary Hickes more confidently than the model does. In most circumstances I would treat that as a red flag. The reason I am not walking away from the selection is that the tie-break rule exists precisely for days like this — when scores are flat across the board, deferring to the market's sharpest opinion is a defensible read rather than a reckless one. The steamer confirms late money arrived; the question is whether the market knows something the feature model has not yet captured.
The negative model edge of -25.3% is the honest headline risk: the market rates Zachary Hickes more strongly than the composite does, which either means the market is sharp and I am following it correctly, or the market is overreacting to the same late money that triggered the steamer signal. I cannot fully distinguish between those two interpretations. Draw data is unavailable for Wolverhampton across the entire shortlist — a known gap in the v2 build — and over 1m5f219y on an AW surface, stall position can matter. Going is recorded as good but Wolverhampton AW rarely varies dramatically; I am not losing sleep there. The more pointed risk is that Koffee And Kale, running earlier at the same track (19:30 vs 20:30 for Zachary Hickes), has the stronger composite at 6.96/10, a larger model edge of +15.1%, and the same steamer signal — readers who want the engine's top score rather than the tie-break selection should note where the top score sits today. Separately, if you are looking for the longest price on the shortlist, Kenough at 5/1 is the engine's value flag — published separately, not today's bet.
Zachary Hickes · 6/5 · 1 unit · MEDIUM
The Fox’s Wire · Bet of the Day
The algorithm pulled five contenders from today’s card. One earned the ticket — the rest are what it beat. When each race settles, the finishes fill in so you can judge the shortlist, not just the selection.
Position only — payout-neutral. A process check, not a P/L. Click any horse to read its profile.
Value pick 0 of 5 won · -5 units at advised prices · since 11 October 2026. Kept separate from the bet’s record above.
